Apple’s operating systems have a long-standing reputation for strong security on both mobile and desktop platforms. Researchers at the cybersecurity firm Calif, however, say they have found a vulnerability in macOS and developed a privilege-escalation method that can expose parts of the system normally kept locked down.
According to The Wall Street Journal, the researchers relied on Anthropic’s Claude Mythos AI tool to identify the vulnerability and to assist in developing a related attack technique. The report states that Mythos located bugs rapidly, although human researchers were still required to complete the exploit development.
The Wall Street Journal also notes that Anthropic considers Mythos sufficiently risky that it is not broadly released by the developer.
Why this matters
The case highlights a key risk posed by advanced AI tools: they can accelerate discovery of software flaws and, potentially, the creation of methods to exploit them. If an AI helps both to find vulnerabilities and to outline steps for exploitation, it could enable faster and more widespread attacks when used by malicious actors.
Apple’s response and next steps
Apple said user security remains “their top priority.” Calif’s researchers have already met with Apple representatives at the company’s headquarters to discuss the security issue. Neither the researchers nor Apple has published technical details; both parties say they will disclose specifics only after Apple issues a patch to fix the vulnerability.
The incident illustrates the dual-use nature of AI tools: they can be leveraged to improve defenses but also create new threats if powerful capabilities are accessed by adversaries.
